Karan Vikyath Veeranna Rupashree

ad3eh1@r.postjobfree.com +1-347-***-**** linkedin.com/in/karanvikyath github.com/KaranVikyath Madison, WI Education

University of Wisconsin, Madison Madison, WI

Master of Science in Electrical and Computer Engineering Dec 2023 Major: Machine Learning

Courses: Learning Based Methods for Computer Vision, Introduction to Optimization, Mathematical Foundations of Machine Learning, Machine Learning, Introduction to Artificial Neural Networks, Matrix Methods in Machine Learning, and Image Processing PES University Bangalore, India

Bachelor of Engineering in Electrical and Electronics Engineering May 2021 Courses: Linear Algebra, Artificial Neural Networks, Introduction to Computing using Python Research Experience

University of Wisconsin - Madison Madison, WI

Research Assistant under Dr. Daniel Pimentel-Alarcon Jun 2023 - Present

• Incorporated subspace clustering, self-expressive layers, and autoencoders to achieve Deep High-Rank Matrix Completion.

• Attained a significant improvement of approximately upto 60% in completion accuracy on real-world datasets (COIL20, EYaleB, ORL) compared to existing algorithms for 10-90% missing data.

• Achieved a remarkable boost of approximately upto 85% in clustering accuracy on the same datasets in comparison to other algorithms.

Industry Experience

Accenture Bengaluru, India

Application Development Associate Jun 2021 - May 2022

• Developed web services to seamlessly integrate 15+ features with the front-end user interface, reducing manual data entry by 70% and improving user experience.

• Incorporated SAP ABAP skills to transition Adobe forms to Webdynpro components, enhancing accessibility and usability. Skills

• Programming Languages: Python, Julia, C Programming, LateX, HTML, CSS, SQL, VHDL, Java, SAP ABAP

• Software and Tools: Tensorflow, Numpy, Pandas, PyTorch, Flask, Docker, OpenCV, Pillow, Google Cloud Platform(GCP), MATLAB, SciLab, WxMaxima, Ki-Cad, Unity, Multisim, Mathematica Project Experience

Exploring the Landscape: A Survey of Story-to-Image Generator Models UW - Madison Code, Report Nov 2023 - Dec 2023

• Achieved a significant FID score of 52.39 with ARLDM on the SSID dataset, outperforming other story-to-image generator models by 5 - 100 points.

• Successfully managed project timelines, completing training for 5 epochs in approximately 6 hours, showcasing efficiency in model development and experimentation.

FCOS: Fully Convolutional One-Stage Object Detector UW - Madison Code, Report Oct 2023 - Nov 2023

• Implemented FCOS for object detection using PASCAL VOC 2007 dataset. Achieved mAP@50 of 52.2% with ResNet-18 after hyperparameter tuning and 56.2% with ResNet-50, surpassing the base model’s 45.4%.

• Extensive model comparisons and hyperparameter tuning contributed to improved efficiency with an inference time reduction from 142.81s (Base Model) to 127.37s.

Comparative Study of an Image to Recipe Generator UW - Madison Report Oct 2022 - Dec 2022

• Developed a recipe ingredient and instruction generation model using Recipe1M+ data and food images as input.

• Vision Transformer (ViT) outperformed ResNet, DenseNet, and VGG16 in encoding food images with a cosine loss of 0.106 and a median rank (MedR) of 14.1.

• Implemented both Transformer and LSTM models to encode text for recipe ingredients and instructions, with the Transformer achieving superior performance, boasting a MedR of 10 compared to LSTM’s 12.0. Picture to Speech Conversion PES University

Report Aug 2019 - Dec 2019

• Implemented a model for converting pictures to speech, enabling the model to audibly identify objects.

• Created a neural model achieving a 78.4% accuracy in labeling CIFAR-10 images.

• Translated these labels into audio representations employing Grapheme and Phoneme. Publications

• K. V. Veeranna Rupashree, S. Baskar, and D. Pimentel-Alarc on, ”Latent Matrix Completion Models,” International Conference on Machine Learning (ICML), 2024 (Under Review)

Contact this candidate